Who Competes for Job Seekers? Visualizing the Labor Market with Glassdoor Data

Employers today compete for talent in the marketplace. But which companies face the most competition for job seekers? And which other employers are they actually competing against? Conventional wisdom is that employers compete for job seekers within the same industry, but that’s changing. Workers are more mobile than ever, often crossing traditional career boundaries and searching for jobs based on many factors, including career growth opportunities, compensation and benefits, culture fit and more, rather than because the jobs are in a specific industry or area. Using Glassdoor data, we have the unique ability to quantify which employers are really competitors for talent. Using patterns in job clicking behavior by Glassdoor users, we can measure a job seeker’s interest in working for specific employers. Two employers are considered labor market competitors if the same user clicks on jobs for both companies. These data offer a unique view into the actual network of labor market competition facing employers—what economists might call the “revealed” labor market.[1] Measuring Competition for Talent To measure which employers are true competitors in the labor market, we created a network of companies based on users’ job clicking behavior. By mapping all of these connections between companies, we can build a rich network showing which employers are considered competitors by job seekers. To measure this competition, we use a method that looks at which companies are centrally located or “influential” in the network of employers, and assign them an influence score. The more influential a company is means they face the most competition, but also have the potential to exert the most influence in the labor market network (by changing employee compensation, for example). The influence score is determined by:

• The number of competitors;
• The number of mutual job clicks between the company and each competitor; and
• The influence score of each competitor.

The score ranges from zero to 1, with a score of 1 given to the company with the most influence on talent competition. Because the influence score measures the influence of a company’s talent competitors, it accounts for the network beyond each employer’s immediate labor market.[2] Who are the most influential employers in today’s job market? Table 1 shows the 20 most influential employers in June 2015. The top five companies were Oracle, IBM, Hewlett-Packard, Accenture and Bank of America. These employers have the largest number of competitors for talent, or have very influential competitors, or both. Table 1. Which Employers are Most Influential in the Labor Market?

Company	Number of Job Market Competitors	Influence Score
Oracle	2,811	1.00
IBM	2,478	0.76
Hewlett-Packard	2,194	0.68
Accenture	2,467	0.63
Bank of America	2,188	0.57
Google	2,143	0.56
Amazon.com	2,149	0.51
Lockheed Martin	1,998	0.41
Microsoft	1,596	0.38
J.P. Morgan	1,691	0.37
Citi	1,704	0.35
Deloitte	1,860	0.32
Apple	1,765	0.31
Target	1,734	0.28
AT&T	1,753	0.27
Aerotek	2,361	0.27
Wells Fargo	1,545	0.26
Lowe's	1,789	0.26
Comcast	1,775	0.26
GE	1,824	0.26

Source: Glassdoor Economic Research A Closer Look While it’s possible to view the entire labor market network, the complexity makes it hard to digest. Instead, we’ll take a closer look at the network of competitors surrounding three big employers today: Oracle, Amazon and Target. The figures below show each of these companies along with their network of direct labor market competitors.[3] The thickness of the links between the companies represents the number of mutual job seekers between them, with thicker links representing closer job market competitors. The size of the company nodes represents the number of job seekers that only click on jobs at that employer. Oracle The most centrally located company in the Glassdoor labor market is Oracle. The above figure shows Oracle’s major competitors for talent. The reason Oracle is so central is not only because it has many competitors, but also because many are influential employers, such as Hewlett-Packard, IBM and Accenture. Included among Oracle’s largest talent competitors are some employers in the same industry like Cisco Systems, but also companies in very different fields such as Bank of America and American Express. The key take-away is that Oracle’s jobs serve as a point of comparison for workers in a wide variety of fields. As a result, any changes in wages, benefits or other HR policies at Oracle may have cascading implications throughout the broader labor market because of its influence. Amazon Amazon ranks as the seventh most influential employer in the job market according to our research. The figure above shows Amazon’s main competitors for talent. Tech firms like Google and Microsoft make up a large part of Amazon’s labor market competition. But there is also a heavy geographic concentration of competitors in the Seattle area, such as T-Mobile, Starbucks, Expedia and University of Washington. This highlights a key feature of the “revealed” labor market: Many tech companies compete for workers regardless of location. From the figure above, we see that Amazon’s job offerings influence Apple and Google, who in turn both influence each other. By contrast, Amazon also competes with Seattle-based employers like Nordstrom and Expedia. But those two companies are mostly isolated from each other, with few job seekers comparing job offers between them. Target Finally, retail giant Target ranks as the 14^th most influential employer in the June 2015 labor market. The figure above shows Target’s main competitors for talent. The labor market facing big retailers differs dramatically from that of tech employers. As expected, big retailers compete with other big retailers for talent, with Lowe’s, Best Buy and Costco appearing as major competitors. However, it’s interesting that companies in home improvement compete among themselves—with Lowe’s, Home Depot and Costco all competing with each other for talent—while Whole Foods, Macy’s and Best Buy have few job seekers in common. This means that whenever Target changes wages or benefits, when its competitors react to remain competitive there will likely be an immediate cascading effect on home improvement retailers than in other retail sectors. Conclusion Which companies are linked with each other in the labor market? In this article, we use a unique data source to identify which companies are competitors for talent: real-time job clicking behavior on Glassdoor. The results offer powerful insights about which employers are considered competitors by job seekers themselves. Two key insights from our analysis are:

• Job seekers don’t always search for positions based on a specific geography, industry or occupation. Instead, they search for a combination of job characteristics that is unique for each job seeker. The mixture of these unique preferences among job seekers combines to create what we call the “revealed” labor market facing companies.

• The “revealed” labor market isn’t equally competitive for companies, and isn’t limited to an employer’s geography or industry. A change in a company’s pay and benefits can have cascading effects throughout the network of labor market connections, and more influential firms are likely to have a wider effect on the job market overall.

Footnotes: [1] Complete details of the analysis are provided in the methodology section below. [2] This measure of network influence is known as “weighted eigenvector centrality,” and is similar to the framework used by Google’s PageRank algorithm (Bryan and Leise 2006). It assigns a score between zero and 1 to each employer, which is proportional to the score of that company’s competitors (with the most influential company receiving a score of 1). See Bryan, Kurt and Leise, Tanya (2006). “The $25,000,000,000 Eigenvector: The Linear Algebra behind Google,” SIAM Review, Volume 48, Issue 3. For complete details of the analysis, see the methodology section below. [3] For the three networks displayed, we restrict to only companies with more than 75 mutual job clickers to simplify the network visualization. Methodology: This research is based on job clicking behavior done by anonymous Glassdoor users on U.S.-based employers’ job openings. We consider a relationship between a user and an employer if users click at least three times on a job posting from the employer within the month of June 2015. Additionally, to be considered within this data set, employers must have at least 30 user relationships within the timeframe. The resulting data set consists of more than 500,000 relationships between more than 3,600 employers and approximately 250,000 users. Employers i and j are considered “linked” if there is at least one user who has a relationship with both companies. As a measure of network centrality, we use weighted eigenvector centrality (WEC). The WEC of firm i is the normalized i-th element of the eigenvector that is associated with the largest eigenvalue of the weighted adjacency matrix of network connections. The matrix is of size n by n where n is the number of firms and the weight in element x_ij is the number of users clicking on i who also click on j. We set x_ii to the number of users that only click on firm i. This is a common measure used when studying influence in networks. For the network visualizations shown, we restrict the network to those links representing at least 75 mutually clicking users for simplicity. The thickness of each the link is proportional to the number of users mutually clicking on the two employers. The size of each node is proportional to the number of users that only click on jobs from that employer.